Your build is green. I measure whether that green means anything.

A gate that does not bite is not a gate. What I look for are checks that report success without having checked anything: a step that goes green on a 404, a suite that crashes before it writes a result, a mutant that no test kills.

• A release step in a Rust project could go green on a 404 and record the SHA of the error response. I reported it; the maintainer wrote and merged a fix within three days — unhappychoice/gitlogue#236

• My own CI once ran 58 test scenarios, crashed before writing any result, and the step still reported success, because continue-on-error hides more than you expect — Do not trust the green checkmark

• The mutants I use to check whether my own gates bite are public, not described — hafiza-kur/faz0

• A documentation patch of mine is in drift's develop branch, committed by the maintainer — simolus3/drift@d9f88e5

• gedik — a read-only security-audit skill for Claude Code. Every finding ships a working PoC; what it cannot measure it reports as not measured, not clean.
• hafiza-kur — a portable project-memory gate system for AI agents. Ships with its own mutants, so the gates can be shown to bite.
• Momentum — offline-first task management, ASP.NET Core and Flutter, with sync and conflict resolution.
• epson-l3251-usb-reset — reading, backing up and resetting a printer's waste-ink counter over IEEE 1284.

23 years in technical acceptance and maintenance of high-tech systems, including auditing a quality management system, but months, not years, in software. My code is mostly written by AI (Claude Code); the design, the checks and the verdicts are mine. I have not worked inside a team codebase or run a live system with real users. What I offer is a measurement, not seniority.

Erzurum, Turkey (UTC+3) · remote only · part-time (10–20 hrs/week), or a fixed-scope audit of a test suite and its CI gates.

Writing · onurkesimbjk@gmail.com